Spatiotemporal persistent homology for dynamic metric spaces (Q2230910)

A dynamic metric space is a pair \(\gamma_X=(X,d_X)\) consisting of a finite set \(X\) and a map \(d_X:{\mathbb R}\times X\times X\rightarrow{\mathbb R}\) such that (i) for each \(t\), \(d_X(t,\cdot,\cdot)\) is a pseudo-metric on \(X\), and (ii) for \(x,x'\), the function \(t\mapsto d_X(t,x,x')\) is continuous, and not identically zero for \(x\neq x'\). Given a closed interval \(I\) and a real number \(\delta\), the authors define a Vietoris-Rips style simplicial complex: the vertex set is \(X\), and a subset \(\sigma\) is a face provided that \(\min_{t\in I}d_X(t,x,x')\leq\delta\) for all \(x,x'\in\sigma\). This gives a collection of simplicial complexes filtered by closed intervals and real numbers, called the spaciotemporal Rips filtration. In a previous article, the authors generalized the Gromov-Hausdorff distance to define the distance \(d_{\mathrm{dyn}}(\gamma_X,\gamma_Y)\) between two dynamic metric spaces. In this article, a polynomial time algorithm is devised to compute a lower bound for \(d_{\mathrm{dyn}}\) that uses the three-parameter persistent homology of the spaciotemporal Rips filtrations of \(\gamma_X\) and \(\gamma_Y\). In particular, both the rank invariant and the zeroth Betti numbers can be used in the algorithm. As a consequence, it is shown that the persistent homology of a spaciotemporal Rips filtration is stable with respect to \(d_{\mathrm{dyn}}\).